# Sign in with Codex

After [connecting chat](/docs/cloud/#connect-your-chat), sign in with the ChatGPT
account you want Codex to use. Your Enso Cloud account does not include model access.

## Connect your account

1. Choose **Get Codex sign-in code** and copy the one-time code.
2. Open **Sign in with ChatGPT** and enter that code on OpenAI's page.
3. Sign in and approve access. If required, enable device-code login in your
   ChatGPT security settings; a workspace administrator may need to allow it.
4. Return to Cloud and choose **Connect Codex** to finish.

Enter the device code on OpenAI's page. Codex does not return a code to paste
into Cloud. Approval and **Connect Codex** are separate steps: approval alone
does not activate the new login in your instance.

The official CLI performs sign-in inside the guest. Provider credentials stay
there; Cloud does not save them or the one-time code in its database or browser storage.

## Resume or retry

An attempt lasts up to fifteen minutes, including the final connection step, and
Cloud shows the exact expiry time beside **Cancel sign-in**. Refresh in the same
Cloud account session to recover the code. Another session may see a busy message.
Continue in the original session or wait for expiry.

Use **Cancel sign-in**, or let an abandoned attempt expire, before starting again;
the next attempt then begins with **Start a new sign-in**. If the final connection
response is lost, let Cloud check status before retrying. Changes to instance
ownership or provider configuration can require a fresh attempt.

**Sign in using the console** is available on every instance and opens by default on
older machines without the guided helper. Use **Open console**, run
`codex login --device-auth` there, or `codex login` with the remote browser if device
sign-in is unavailable for your account, then choose **Check sign-in**. Follow
[manual setup](/docs/cloud/#manual-setup) for older instances without guided setup.

## Verify a reply

Choose **Start Enso** and send a normal message in the paired chat. The setup
check requires a successful reply from the selected provider and model. A stored
login alone does not complete setup or prove model access.